A Heterogeneous Windows Cluster System for Medical Image Reconstruction

This paper demonstrates a heterogeneous windows cluster system to obtain high performance computing (HPC) resources for medical image reconstruction applications. The communication in the cluster system is powered by MPICH open source library. Our works are to provide a simple processes management (PM) system with the functions of load balance in the windows version 1.25 of MPICH library and some PM extensions in the windows version 1.03 of MPICH2 library. These extensions ensure our medical image reconstruction applications can smoothly and robustly be running on a heterogeneous windows network environment, which is considered as a windows MPI cluster system with a low rate of cost/performance. In the study cases, the OSEM algorithm and Katsevich algorithm are implemented on our windows cluster system. The experiment results in a broad variety of image volumes are presented to verify system's feasibility and applicability to increase speedup of HPC powered CT image reconstruction